- references
- goals of this workshop
- workshop plan
- implement
ZIOCustom.serviceWithZIOto inferR,E,Atypes- it may be valuable to first take a look at: https://github.com/mtumilowicz/scala-cats-functional-dependency-injection-workshop#izumi-tag
- implement
TransactionService.inTransactionto inferR,E,Atypes
- implement
- good analogy to set intuition on partially applied types
- example
- problem
def map[A, B](list: List[A], f: A => B): List[B] = list.map(f) val nums = List(1,2,3) mapping function(nums, _ + 1) // not compiling mapping function(nums, (i: Int) => i + 1) // compiling - solution - partial application of arguments
def map[A, B](list: List[A])(f: A => B): List[B] = list.map(f) val nums = List(1,2,3) mapping function(nums)(_ + 1)
- problem
- in scala, you can specify either all types or none params
- example
- problem
def serviceWithZIO[Service: Tag, R, E, A](f: Service => ZIO[R, E, A]): ZIO[R with Service, E, A] = ZIO.service[Service].flatMap(service => f(service))ServiceandR,E,Aare not connected -Servicehas to be specified, rest can be inferred
- solution - partial application of types
def serviceWithZIO2[Service] = new ServiceWithZIOPartiallyApplied[Service] final class ServiceWithZIOPartiallyApplied[Service]() { def apply[R, E, A](f: Service => ZIO[R, E, A])(implicit tag: Tag[Service]): ZIO[R with Service, E, A] = ZIO.service[Service].flatMap(service => f(service)) // tag is needed by ZIO.service[Service] }
- problem
- usually used for wrappers, where we have groups of unconnected types
- performance
- to keep it at zero cost, we have to use 'AnyVal'
final class ServiceWithZIOPartiallyApplied[Service](private val dummy: Boolean = true) extends AnyVal { ... }
- to keep it at zero cost, we have to use 'AnyVal'